Field notes
How to read peak service-request hours without overreacting
Operators often react to the loudest day of the month. A festival weekend, a school holiday, or a viral promotion can look like a permanent shift in demand. For service request analytics on on-demand apps, the useful question is whether that peak repeats in the same hour band across several ordinary weeks.
Start by lining up request timestamps for at least three non-event weeks. Mark national holidays and your own campaign dates so they sit outside the baseline. Then count requests by hour within each coverage zone, not only company-wide. A city-wide evening spike can hide a midday concentration in one district that actually drives missed acceptances.
Treat a peak as durable when it appears in the same window across most baseline weeks and when fulfilment lag rises with it. Treat it as reactive when volume jumps once and lag stays flat because capacity already covered the surge. Roster changes belong to durable peaks; temporary overtime or a borrowed float belongs to reactive ones.
When we run Demand Window Mapping at Service Pulse Core, we hand back a short table of candidate windows with a confidence note based on how many clean weeks were available. That keeps the conversation on evidence rather than the memory of last Friday’s scramble.